INTEL LLM: Neural-chat-7b vs PPLX-7B-online

INTEL released a new LLM, trained on Gaudi2, on an open Apache License 2.0. Current version is Neural-chat-v3-1, available on HuggingFace.

But is it better than the brans new perplexity pplx-7b-online, with the ability to search the internet? Or is it comparable to the pplx-7b-chat?

Since it is based on an Mistral-7b, it all depends on the specific fine-tuning (a synthetic GPT-4 generated dataset) by Intel and the DPO Reinforcement Learning.

#reasoning
#ai
#aieducation

apache chat current gaudi2 huggingface intel internet license llm mistral perplexity pplx search